# How agents, skills, and data work together in AWS Quick

- Source: MechNext — https://mechnext.eu/en/article/67-how-agents-skills-and-data-work-together-in-aws-quick
- Published: 2026-11-10
- Updated: 2026-11-10
- Category: Industry News
- Language: en (Markdown in other languages: ?lang=de, ?lang=sk, ?lang=pl, ?lang=zh, ?lang=ru)

> AWS Quick combines configurable AI agents with knowledge sources, connectors, and reusable skills. This overview explains each component, the user-based access model, and why publishing and sharing are separate steps.

AWS Quick is an AI-powered service for automating tasks, analyzing data, building web applications, and conducting research through natural-language interaction. An agent is one of its central building blocks: a configurable AI with its own instructions, knowledge sources, and connectors. It can work interactively in a chat or operate independently through a routine.

The components around an agent serve distinct purposes. The agent defines how a job should be approached and which configured resources belong to that job. Spaces and knowledge bases supply information, connectors link Quick to applications or native tools, and skills capture reusable instructions for specific tasks. Together, these elements turn an agent configuration into a defined combination of guidance, information, and permitted capabilities.

### The built-in Quick agent and custom agents

Quick is the default built-in agent available to every user. In each conversation, it works with the resources that the current user can access. The resources available in practice are therefore determined not only by the agent but also by the access held by the person using it.

A custom agent can be configured for a specific job. Its configuration includes a name, description, instructions, spaces, connectors, and skills. It can be limited to selected spaces, connectors, and skills, then tested and published before being shared with users and groups. This makes it possible to define a focused working context around a particular task.

Whether built in or custom, an agent runs with the access of the person using it. Consequently, the same shared agent can reach different data and skills for different users. Sharing an agent configuration does not by itself give every recipient identical access. The user’s own access remains part of every interaction.

### Spaces and knowledge bases

A space is a curated collection of documents and AWS Quick resources that an agent uses as an information source. It provides a way to gather relevant material into a selected collection for the agent. A knowledge base is an organized, indexed collection optimized for retrieval and question answering. Quick synchronizes supported knowledge bases on a schedule.

The two concepts describe ways of supplying organized information, but the stated emphasis differs. A space is a curated collection of documents and Quick resources, while a knowledge base is indexed and optimized for retrieval and answers. Both may contribute information to an agent without changing the underlying rule that the agent operates with the current user’s access.

### Connectors connect data and actions

A connector links AWS Quick to an external application or a native tool. Through that connection, agents can read data and take actions. What an agent can reach through a connector is still governed by the access of the person using the agent.

There is also an operational distinction between cloud and local connectors. Cloud connectors are available across devices. A local connector depends on its host machine: that computer must be on and connected for the connector to be available. This distinction matters when a configured agent is expected to work from different devices or when a routine depends on a locally hosted connection.

### Skills package reusable working instructions

A skill is a reusable set of instructions that teaches an agent how to perform a specific task. Quick skills follow the Agent Skills specification. Structurally, a skill consists of a folder with a required root-level `SKILL.md` file. The folder may also include optional scripts and reference files. The required file provides the core skill instructions, while the optional materials can support the task described by those instructions.

Quick uses progressive disclosure when handling skills. Initially, it reads the name and description of each skill. It loads the complete instructions only when a request calls for that skill. The short identifying information therefore helps Quick determine relevance, while the full instructions are reserved for the point at which the particular workflow is needed.

This separation also clarifies the relationship between agents and skills. An agent is the broader configuration: it has its own identity, instructions, knowledge sources, connectors, and selected skills. A skill is a reusable task-specific instruction set within that environment. Restricting a custom agent to selected skills defines which packaged procedures are part of its configured job, subject to the access available to the user.

### Publishing is not the same as sharing

Publishing and sharing a skill are separate steps. Publishing a skill does not, on its own, grant access to other people. If others are meant to use it, the skill must also be shared. Once copies are shared, they remain synchronized when the publisher republishes the skill.

The distinction reflects a broader theme in the Quick model: configuration, distribution, and access are related but not interchangeable. A skill can be published without being shared. An agent can be shared while producing different reachable resources for different users. A connector can be configured yet depend on the relevant user access and, for a local connector, an active host machine.

> The practical scope of an agent comes from both its configuration and the access of the person currently using it.

### Reading the complete model

Each component answers a different question. The agent’s instructions describe how it should work. A space supplies a curated collection of documents and Quick resources. A knowledge base provides an indexed collection optimized for retrieval and question answering. A connector opens a path to read data or take actions in an application or tool. A skill supplies a reusable procedure for a defined task.

These components can be assembled around the default Quick agent or a custom agent designed for a specific job. Testing and publication prepare a custom agent for sharing, while user and group sharing determines who receives it. Even after that point, each person uses it through their own access. Understanding that boundary is essential when considering why two people using the same shared agent may not reach the same data or skills.

Viewed as a whole, AWS Quick combines natural-language interaction with configurable agents and clearly separated supporting resources. The agent brings instructions and selections together; spaces and knowledge bases organize information; connectors provide links to data and actions; and skills preserve reusable task instructions. Publishing and sharing then govern distribution, while the current user’s access continues to govern what the agent can actually use.
